When Should a Research Agent Stop Searching? A Founder's Three-Part Stopping Rule
Pitiful-Minute-2818 · reddit · 2026-10-18
A Reddit discussion on the hardest part of web-research agents: deciding when there's enough information. The poster (founder of Dreamtune Labs) splits the stopping decision into three checks:
- Coverage: is there evidence for every requirement, including reasons the recommendation might be wrong?
- Progress: are recent searches adding new evidence, or just variations of what's already known — ten pages repeating the same source isn't progress
- Budget: has a limit on time, spend, or tool calls been hit?
Key point: hitting a budget limit doesn't mean the research is complete — output should state which questions remain unresolved rather than present a confident answer just because execution stopped. He advocates defining evidence requirements upfront and returning either a supported answer or a clear account of what's missing, and asks what stopping rules other agent builders actually use.
More from coding & agent
- DeepSeek harness lifts V4.1 Flash by 2-3 points at 80% higher cost; model caught cheating — PawelHuryn · 2026-10-18
- Agent engineering debate: when is a retry defensible after a timed-out non-idempotent write? — InformationLive493 · 2026-10-18
- Open-source Unithread unifies Node and browser multithreading behind one API — jasonkneen · 2026-10-18
- Pi Is More Than a CLI: A Map of Its Core Libraries for Building Your Own Agent — solyarisoftware · 2026-10-18
- APEX-MEM: property graph + append-only events for LLM long-term memory — burkov · 2026-10-18
- Agently demos macOS agent that runs email, invoicing and CRM workflows from one command — Scobleizer · 2026-10-18